Mix

From APL Wiki
Revision as of 23:49, 29 September 2020 by Marshall (talk | contribs) (→‎Language support: Add BQN)
Jump to navigation Jump to search

Mix ( or in nested APLs; Disclose or Open > in flat array languages) is a primitive monadic function which reduces the depth of its argument by combining all its element arrays into a single array. The shape of its result is then the shape of the argument followed by the common shape of the argument elements. Depending on the definition of Mix, the elements may be required to share the same shape, or they may be extended in shape (with fill elements) or rank to fit a common shape. In all APLs Mix allows a simple array argument, and returns it unchanged. Mix is a left inverse to both Enclose and Split, while the function First, which is identical to Mix on scalars, is a left inverse to Enclose only.

History

Early experiments in array nesting usually defined a primitive to extract the contents of a nested scalar, but did not always specify how it should behave on an array. Jim Brown's "A Generalization of APL",[1] which would become APL2, defined such a primitive, "reveal", while stating that it "may be undefined" on non-scalar arrays.

Mix () was first introduced by NARS in 1981.[2] It required all elements to have matching shape, and allowed a specified axis to define where in the result array their axes would be placed. In the same year, SHARP APL introduced the concept of a boxed array, along with the function Disclose (>).[3] In SHARP, Disclose was simply defined to be the inverse of Enclose, but given function rank 0 so that it would disclose each box in an argument and combine the results. Based on SHARP's definition of Rank, the arrays to be combined were required to have the same shape, and because of its flat array model, they also had to have the same type (numeric, character, or boxed), as otherwise the result array would not be representable.

APL2, released in 1984, used the same name Disclose as SHARP APL but introduced the glyph . It shared NARS's function axis definition, but extended the function to allow argument elements with different shapes as long as they had the same rank. Arrays with different shapes would be padded using fill elements as with Take to bring them to a common shape large enough to contain every element.

A further extension was introduced as a consequence of J's definition of the Rank operator. In J, result arrays in a function applied with rank can have different ranks, and results with lower rank are brought to a common rank by adding leading 1s to the shape. This change was taken up by Dyalog APL in version 14.0, which introduced Rank following J's definition, in order to make Mix and Rank consistent. However, Dyalog differs from J in that it permits mixed arrays, so while in J an array containing a mix of numbers, characters, and boxes cannot be mixed, in Dyalog any array at all can be mixed.

Language support

Mix is only available in languages which allow nesting, either by boxes or nested arrays, so it is not present in early APLs such as APL\360 and APL.SV.

Languages Name Glyph Requirements NARS-style function axis APL2-style function axis
NARS Mix Shape Yes No
SHARP APL, A+ Disclose > Shape and type No No
APL2, APLX Disclose Rank No Yes
NARS2000 Disclose None No No
Dyalog APL Mix or (ML) None Yes Yes
J Open > Type No No
ngn/apl Mix None No No
dzaima/APL Merge Rank No No
BQN Merge > Shape No No

External links

Lessons

Documentation

References

  1. Jim Brown. "A Generalization of APL" (Ph.D. thesis). 1971.
  2. Carl M. Cheney. APL*PLUS Nested Arrays System (NARS reference manual). STSC. 1981.
  3. Ken Iverson. SATN-41: Composition and Enclosure. IPSA. 1981-06-20.


APL built-ins [edit]
Primitives (Timeline) Functions
Scalar
Monadic ConjugateNegateSignumReciprocalMagnitudeExponentialNatural LogarithmFloorCeilingFactorialNotPi TimesRollTypeImaginarySquare RootRound
Dyadic AddSubtractTimesDivideResiduePowerLogarithmMinimumMaximumBinomialComparison functionsBoolean functions (And, Or, Nand, Nor) ∙ GCDLCMCircularComplexRoot
Non-Scalar
Structural ShapeReshapeTallyDepthRavelEnlistTableCatenateReverseRotateTransposeRazeMixSplitEncloseNestCut (K)PairLinkPartitioned EnclosePartition
Selection FirstPickTakeDropUniqueIdentityStopSelectReplicateExpandSet functions (IntersectionUnionWithout) ∙ Bracket indexingIndexCartesian ProductSort
Selector Index generatorGradeIndex OfInterval IndexIndicesDealPrefix and suffix vectors
Computational MatchNot MatchMembershipFindNub SieveEncodeDecodeMatrix InverseMatrix DivideFormatExecuteMaterialiseRange
Operators Monadic EachCommuteConstantReplicateExpandReduceWindowed ReduceScanOuter ProductKeyI-BeamSpawnFunction axisIdentity (Null, Ident)
Dyadic BindCompositions (Compose, Reverse Compose, Beside, Withe, Atop, Over) ∙ Inner ProductDeterminantPowerAtUnderRankDepthVariantStencilCutDirect definition (operator)Identity (Lev, Dex)
Quad names Index originComparison toleranceMigration levelAtomic vector